RFR: 8065138 - Encodings.isRecognizedEnconding sometimes fails to recognize 'UTF8'
huizhe wang
huizhe.wang at oracle.com
Thu Nov 20 01:07:26 UTC 2014
On 11/19/2014 4:09 PM, Mandy Chung wrote:
>
> On 11/19/2014 3:49 PM, Mandy Chung wrote:
>>
>> On 11/19/2014 12:50 PM, Daniel Fuchs wrote:
>>> On 11/19/14 9:36 PM, Mandy Chung wrote:
>>>> resources.jar will be gone when we move to the modular runtime
>>>> image (JEP 220 [1]).
>>>> JDK-8065138 and JDK-8065365 will become non-issue in JDK 9.
>>> Do you mean that the property files will no longer be stripped of
>>> their comments?
>>
>> (sorry for my delay in reply as I was trying to get the number of the
>> resources in the modular image vs resources.jar but got distracted.)
>>
>> The current version copies all bytes when generating the modular
>> image. This is a good question whether we should strip off the
>> comments when writing to the modular runtime image. I think we
>> should look at the footprint and any performance saving and determine
>> if we should do the same in JDK 9.
>>
>
> I looked at the exploded image under BUILD_OUTPUTDIR/jdk/modules/java.xml
> and found that the comments of Encodings.properties are stripped. I was
> confused with the mention of resources.jar that I assume it was a step
> stripping the comments before writing to resources.jar. This is still
> an issue in jigsaw m2 I believe.
>
> Where does the build strip the comments?
A previous issue was that the build process strips off anything after
'#' when copying properties files. In JDK8:
jaxp/make/BuildJaxp.gmk:
$(JAXP_OUTPUTDIR)/classes/%.properties: $(JAXP_TOPDIR)/src/%.properties
$(MKDIR) -p $(@D)
$(RM) $@ $@.tmp
$(CAT) $< | LANG=C $(NAWK) '{ sub(/#.*$$/,"#"); print }' > $@.tmp
$(MV) $@.tmp $@
This was fixed in JDK 9. The per-repo process was removed. It looks
like the properties processing process is now consolidated into
make/common/JavaCompilation.gmk. So the issue Daniel found is new (in
terms of stripping). Search 'properties files' to see the macro.
Joe
>
> Mandy
>
More information about the core-libs-dev
mailing list